Unusual transaction amount rule

An unusual transaction amount rule detects a payment whose size is out of line with the account's own history, rather than with a fixed limit applied to everyone. It compares each amount with what is normal for this particular account, which lets it flag a modest payment on a low-value account and ignore a large one on a business account that routinely moves it.

The fraud scenario this rule addresses

A fraudster does not need to move a large absolute amount to cause serious harm. An account whose payments are normally small, drained with a payment several times its usual size, is a loss even though the figure would pass any fixed limit. The deviation from the account's own pattern is the signal.

The reverse also holds: a fixed limit set low enough to catch the first case would alert constantly on accounts where large payments are ordinary.

Data the rule requires

Example logic

Illustrative pseudocode: it compares each event with the rule's parameters and flags the ones that match. Parameters are symbolic: every value is institution-specific and has to be derived from your own transaction distribution.

RULE unusual_transaction_amount
  FOR EACH event E on account A:
    history = amounts of A's transactions of type E.type
              WHERE timestamp > E.timestamp - :lookback
    IF count(history) < :min_history:
      RETURN                            # too little history to judge
    typical = median(history)
    spread  = median_absolute_deviation(history)
    deviation = (E.amount - typical) / max(spread, :spread_floor)
    IF deviation >= :deviation_limit:
      RAISE alert(rule = "unusual_transaction_amount",
                  account = A, evidence = E + summary(history))

Choosing a threshold

This rule has more parameters than a fixed-limit rule, and they pull against each other. The look-back period decides how much history defines normal: short and the baseline tracks recent behaviour but is easily pulled by a single unusual month, long and it is stable but slow to recognise a genuine change in circumstances.

The deviation limit is the sensitivity. A low limit flags ordinary variation, and a high one only the extreme. The spread measure matters because an account whose payments are all nearly identical has almost no spread, so any different amount appears extreme; a spread floor stops that from alerting on a one-unit difference.

The minimum history is the guard for new accounts. Without it the rule is unreliable on new customers, and with a high minimum a new account has no amount monitoring at all during the period when mule accounts are most active. Pairing it with a fixed-limit rule covers that gap.

False positives to expect

Legitimate amounts change when life does: a first rent payment after moving, a wedding, a deposit on a home or car, an annual insurance or tuition payment, a seasonal business whose volumes jump at certain times, a pay rise, or an inheritance. The customer's behaviour has changed for a real reason, not a fraudulent one.

Recurring large payments are often predictable, so recognising them as part of the account's pattern, or treating a payment to a long-standing payee differently from one to a new payee, removes much of the noise.

Investigating an alert from this rule

  1. Review the account's amount history and confirm the flagged payment really is out of pattern, and by how much.
  2. Check the type and destination. A large payment to an established payee is a different case from one to a new payee.
  3. Check for a change in circumstances that explains it, such as a recent salary change, a deposit that arrived, or a known annual payment.
  4. Check the device, channel and location against the account's usual behaviour.
  5. Look for other rules that fired on the account around the same time, since a combination is stronger than any single alert.
  6. Contact the customer if the evidence remains unclear, and record the explanation so the account's baseline reflects the new normal.

Related rules

Where this rule fits

The use case this rule supports is covered in unusual transaction behaviour, and the fraud rules engine page describes how conditions like these are expressed and versioned.

Common questions

What is an unusual transaction amount rule?

An unusual transaction amount rule is a fraud rule that compares a payment with the typical amounts on the same account and raises an alert when it falls well outside them. Unlike a fixed limit applied to everyone, it uses the account's own history as the reference, so it can flag a modest payment on a low-value account and ignore a large one on a business account.

The rule depends on enough history to define normal. For a new account there is too little, so the rule usually stays quiet there and is paired with a fixed limit to cover the gap.

Its main tuning decisions are the look-back period and the sensitivity. Genuine life events such as a house deposit or an annual payment create deviations that are legitimate, so recognising predictable large payments reduces noise.

Discuss your transaction monitoring requirements

Tell us which transaction flows you need to monitor, which systems Akoben would integrate with, and how your fraud team works today.

Request a demo